# NVIDIA: Parakeet TDT 0.6B v3

Canonical model ID: `nvidia/parakeet-tdt-0.6b-v3`

Fast multilingual speech-to-text for 25 European languages including Czech, Slovak, German, Polish and English, with automatic punctuation, segment and word timestamps, and SRT/VTT output. Audio up to 60 minutes per request. Model by NVIDIA, licensed under CC-BY-4.0.

## Runtime metadata

This table is generated from public model metadata. Readiness and capacity are refreshed from `GET /v1/models` on page load.

| Property | Value |
|---|---|
| Kind | `audio-transcription` |
| Input | audio |
| Output | text, transcription |
| Context | Not token-based |
| Maximum output | Endpoint-specific |
| Supported endpoints | `/v1/audio/transcriptions` |
| Capabilities | Endpoint-specific |
| Pricing | $0.0014 per audio minute |

## Use this model

Copy the canonical ID into the `model` field. Check the endpoint guide before using optional capabilities.